Statistical models will be constructed for analyzing geographic variations in mortality rates of different types of cancer. New methods will be developed for simultaneously estimating mortality rates for a large number of geographic units, detecting units with unusually high or low rates and testing for the presence of environmental effects which explain geographic variations. Cancer mortality data for Missouri cities during 1972-76 are being analyzed for possible relationship with municipal drinking water supplies. This will be used as an illustration for relating other environmental variations in mortality and morbidity in Missouri and in other geographical areas.